LogicLink: Version 5

LogicLink is a conversational AI chatbot developed by Kratu Gautam (AIML Engineer). Powered by the TinyLlama-1.1B-Chat-v1.0 model, it provides an interactive interface for engaging conversations, query resolution, and task assistance. Version 5 features streaming responses, conversation management, and a sleek GUI.

---

🔍 Topic Index

• ✨ Key Features
• 📸 GUI Display
• 🛠️ Installation
  • Prerequisites
  • Setup
  • Directory Structure
• 💬 Usage
• ⚙️ Technical Architecture
  • Model Configuration
  • Key Components
• 🧪 Troubleshooting Guide
• 🚀 Future Roadmap
• 📜 License

---

✨ Key Features

Feature	Description	Benefit
🤖 Conversational AI	TinyLlama-1.1B-Chat-v1.0 powered responses	Natural, engaging dialogue
⚡ Streaming Responses	Real-time token generation with TextIteratorStreamer	Smooth user experience
🎨 Customizable GUI	Red/blue/black theme with Gradio & ModelScope Studio	Professional interface
🗂️ Conversation Management	New chat, clear history, delete conversations	Full control over interactions
⏱️ Single Time Stamp	Regex-cleaned response timing *(4.50s)*	Consistent performance metrics
🚀 CUDA Support	Automatic GPU detection with CPU fallback	Optimized performance
🛡️ Error Handling	Graceful failure for memory/input issues	Robust user experience

---

📸 GUI Display

---

💬 Full-Fledged Conversation

LogicLink engaging in a complete dialogue, handling multiple turns seamlessly.
This demonstrates its ability to maintain context, respond naturally, and adapt to user intent across an extended session.

---

🧑‍💻 Coding Response (Part 1)

LogicLink generating a structured coding solution.
Notice how it explains the reasoning step-by-step, making the output not just correct but also educational.

---

🧑‍💻 Coding Response (Part 2)

A continuation of the coding workflow, where LogicLink refines and expands on its earlier solution.
This shows its iterative reasoning ability — improving code quality when prompted.

---

🔑 Core Response

A snapshot of LogicLink delivering a core logical explanation.
This highlights its strength in breaking down abstract queries into clear, actionable insights.

---

⚡ While Processing

The system mid‑inference, showing its real-time feedback loop.
This reassures users that LogicLink is actively working on their request.

---

🔄 With vs Without Latest Output Text Box

A side‑by‑side comparison of LogicLink’s performance with and without LOTB (Latest Output Text Box).
The difference illustrates how LOTB enhances reasoning depth and response clarity.

---

📊 Bottom Section

The footer view of the interface, where conversation summaries and quick actions are displayed.
This ties the user experience together, making LogicLink feel like a polished, end‑to‑end assistant.

---

🛠️ Installation

Prerequisites

• Python 3.8+
• CUDA-enabled GPU (recommended)
• Dependencies:

  pip install gradio torch transformers modelscope-studio

---

Setup

1. Clone repository:

   git clone https://github.com/Kratugautam99/LogicLink-Project.git
   cd LogicLink-Project

2. Install dependencies:

   pip install -r requirements.txt

3. Run application:

   python app.py

---

Directory Structure

LogicLink-Project/
├── LogicLinkVersion5.ipynb
├── README.md
├── app.py
├── config.py
├── .gitattributes
├── requirements.txt
├── assets/
├── Documents/
├── Screenshots/
├── ui_components/
└── Different Versions of LogicLink/  (not expanded)

---

💬 Usage

# Sample interaction flow
user >> "Who are you?"
LogicLink >> "I'm LogicLink V5, created by Kratu Gautam. How can I assist you today? *(4.50s)*"

1. Interface Controls:

   • 💬 Input field: Type queries
   • ➕ New Chat: Start fresh conversation
   • 🧹 Clear History: Reset current chat
   • 🗑️ Delete: Remove conversations from sidebar

2. Performance Metrics:

   • ⏱️ Response time: 3-5s (GPU), 5-8s (CPU)
   • 💾 RAM usage: 2-3GB (CPU), ~1.5GB (GPU)

---

Technical Architecture

Model Configuration

# Core model parameters
model = AutoModelForCausalLM.from_pretrained(
    "TinyLlama/TinyLlama-1.1B-Chat-v1.0",
    torch_dtype=torch.float16 if cuda else torch.float32
)

# Generation settings
generation_kwargs = {
    "max_new_tokens": 1024,
    "temperature": 0.7,
    "top_k": 50,
    "top_p": 0.95,
    "num_beams": 1
}

---

Key Components

1. Prompt Engineering:

   <|system|>You are LogicLink V5 created by Kratu Gautam</s>
   <|user|>{user_input}</s>
   <|assistant|>
   

2. Streaming Pipeline:

   graph LR
   A[User Input] --> B(Tokenizer)
   B --> C{TextIteratorStreamer}
   C --> D[Model Generation]
   D --> E[Real-time Output]
   E --> F[Regex Cleaner]
   F --> G[Timestamp Append]
   

   Loading

3. GUI Components:

   • pro.Chatbot: Conversation display
   • antdx.Sender: Input field
   • antdx.Conversations: Sidebar manager
   • antd.Button: Action controls

---

🧪 Troubleshooting Guide

Issue	Solution
Double timestamps	Verify regex: re.sub(r'\*\(\d+\.\d+s\)\*', '', response)
Slow responses	Enable CUDA, reduce max_new_tokens to 512
GUI rendering issues	Update packages: pip install --upgrade gradio modelscope-studio
Delete button failure	Check menu_click event binding in JS
Model loading errors	Validate RAM ≥3GB, test with minimal example

Minimal Test Script:

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
model = AutoModelForCausalLM.from_pretrained("TinyLlama/TinyLlama-1.1B-Chat-v1.0")
inputs = tokenizer(["Test input"], return_tensors="pt")
outputs = model.generate(**inputs, max_new_tokens=10)
print(tokenizer.decode(outputs[0]))

---

🚀 Future Roadmap

• Persistent Storage: SQLite conversation history
• Multimodal Support: Image/text inputs
• Enhanced Prompting: Context-aware responses
• Deployment Options: Docker containerization
• Performance: Quantization for CPU optimization

---

📜 License

MIT License - See LICENSE

---

Developed with 🧠 by Kratu Gautam | AIML Engineer
GitHub | HFT Space | UI Framework